Crop Adaptation to Climate Change - CERE- Centre for ... · Crop Adaptation to Climate Change 1 Introduction Crop yields are forecasted to decrease by 30-46% before the end of the

Crop Adaptation to Climate Change Lunyu Xie, Sarah Lewis, Maximilian Auffhammer, Peter Berck1

Abstract Farmers may adapt to climate change by growing different crops. This type of

adaptation may offset the negative effects of climate change on crop yields. However, adaptation

may be restricted by soil conditions. Even in the case of substantial warming the actual amount

of adaptation could be small. In this paper, we pair a 10-year panel of satellite-based crop

coverage in the Midwest with spatially explicit soil data and a fine-scale weather data set.

Combining a proportion type model with local regressions, we simultaneously address the

econometric issues of proportion dependent variables and spatial correlation of unobserved

factors. Based on the estimates of crop choice, we predict the future crop distribution under

several climate change scenarios. We find that rice and cotton spread northward, the average

shares of corn and soy decrease in the north and increase in the south. We also find that crop

shifting patterns vary across quality levels of soils. There is less crop adaptation on better soils

than on soils with lower quality.

1 Lunyue Xie is Associate Professor at Renmin University, Sarah Lewis is a GIS consultant at AMKO consulting; Maximilian Auffhammer is George M. Pardee, Jr. Family Chair, and Peter Berck is S.J. Hall Professor at the College of Natural Resources, University of California at Berkeley. We are grateful to Michael Roberts and Wolfram Schlenker for sharing both their weather data and expertise. This project was funded by the Environmental Biosciences Institute at Berkeley and Illinois. The remaining errors are those of the authors.

Crop Adaptation to Climate Change 1 Introduction Crop yields are forecasted to decrease by 30-46% before the end of the century even under

the slowest (B1) climate warming scenario (Schlenker and Roberts 2009). Farmers may adapt

to the expected yield changes by growing crops more suited to the new climate. Predicting

adaptation behavior, i.e. the change in cropping patterns, is therefore an important part of

evaluating the effect of climate change on food and fiber production. In this paper, we look at

the potential adaptation to climate change, using currently grown crops, for a group of US

states situated in a north-south transect along the Mississippi-Missouri river system. Together

the states comprise a major agricultural region with a considerable diversity of weather and

soil. The selected states are also among the few for which more than 10 years of fine scale

satellite-based crop coverage data are available.

Along the Mississippi River, the dominant crop types are corn, soy, cotton and rice in

the south and corn and soy in the colder north. Based on temperature alone, adaptation to

higher temperatures should result in the northward spread of cotton and rice and substitution

of shorter-season crops (e.g., soy) for longer-season crops (e.g., corn). However, agricultural

crop coverage is not determined by temperature alone, or even rainfall and temperature taken

together. Soil properties are a major determinant of which crops can be grown and what the

crop’s ultimate yield is. It is very plausible that, even in the face of the same level of

warming, crop shifting patterns will be very different across soils of different qualities. The

purpose of this paper is to show how weather and soil determine crop location and how, in

the face of warmer weather, crop adaptation varies across quality levels of soil.

Modern econometric studies of crop land coverage began with Nerlove’s (1956)

examination of crop share response to crop prices. His estimating equations are of the form

that coverage is a function of lagged coverage, crop price, input prices and other variables.

There are many ways to elaborate on this basic model. (1) In many countries (e.g., the United

States and European Union), the incentive to grow crops in addition to the price is

government payments. As these programs change year to year and have different marginal

effects for different farmers, it is not possible to have a fully satisfactory treatment of the

price variable. If the focus, as was Nerlove’s focus, is on price response, parsing the true

incentive effects is a serious problem. If the focus is on climate, the standard solution is to

use year fixed effects to account for both prices and government programs. The year fixed

effects also would account for differences in input prices. (2) Many authors (Just 1974,

Chavas and Holt 1990, Lin and Dismukes 2007) think that the risk of growing a crop, perhaps

the variance or lower semi-variance, is an important determinant of crop choices. So long as

the risk of growing a crop is taken as constant, which is a good approximation in a short time

series, crop fixed effects account for this factor. (3) Crop coverages are proportions and

therefore should sum to one. Indeed there are authors (Lichtenberg 1989, Wu and Segerson

1995) who use discrete choice models for crop share decisions. Berry’s logit (1994) is an

appealing discrete choice model for shares that are not zero or one, because it is linear in the

parameters and errors. However, our crop coverage data has many data points with zero

coverage. To deal with a great number of zero shares, we use a limited dependent variable

regression with a transformation function which will be described in detail later. (4)

Land use can be correlated across space, either as a spatial lag or spatial error process. A

spatial lag process is natural in housing developments, where the land use on the next plot

does influence that on the subject plot. A spatial error process corresponds to unmeasured

factors that influence yield that vary slowly over space. For instance, the persistence of

summer fog along a coast might not be captured by measured low temperature, since the fog

also produced moisture. All adjacent coastal areas would be similarly affected. The spatial

correlation in errors does not cause inconsistency to OLS estimate, but for nonlinear

regressions, such as logit, tobit, etc., the key problem is that the homoscedasticity assumption

is violated by the spatial correlation in errors. We use a local regression framework so that all

parameters and the variances can vary across the landscape and account for unmeasured place

specific phenomenon. (5) Crop coverage in cross section is determined by climate and soil.

Studies interested in price response use place fixed effects to account for these factors.

Because these are the factors that interest us, we use measured climate and soil variables. We

also include the interaction terms of moisture and heat, because a dry warming is likely to be

more harmful than warming with moisture.

Recent literature, particularly Schlenker and Roberts (2009), working at the county

level, quantified the effects of weather on yield. Their work is noteworthy for the use of a

great deal of spatial and temporal detail in their weather data. In their studies, the effects of

soil are subsumed in the place fixed effects. The general tenor of their results is that high

temperatures are very harmful to yields and so climate change projections for the United

States result in large yield deficits in response to an increase in the number of hours of 29°c

plus temperatures for corn, 30°c for soybeans, and 32°c for cotton. Lobel et al. (2011) did a

very similar analysis for Africa, though they emphasize the interaction of moisture and heat.

These studies find that warmer climates negatively affect yield.

One potential response of farmers to climate change is to shift the location of crops, in

turn planting crops with characteristics better matching the new landscape characteristics.

This type of adaptation is evident on crop landscape maps. One sees cotton in the warmer,

wetter south, wheat in drier regions, corn in the wetter parts of the Midwest, and so on. The

choice of crops to fit climate may offset the negative effects of increasing temperature on

crop yields, but it may be limited by soil conditions. Where crops of a certain type can be

grown and what their maximum potential yields may be are determined not only by weather

but also by soils. If all crops suitable to local soils are negatively affected by warming, it is

possible that farmers are left with no better crop to substitute to. That is, adaptation happens

only when the substitution crop fits in the local soils and the current crop is harmed so much

that it is less profitable than the substitution crop. In this paper, we show that, due to soil

restriction, adaptation makes some difference, but that it does not undo the negative effects of

higher temperature.

The remainder of the paper is organized as follows. Section 2 summarizes the data on

land use, soil conditions, weather, and climate change scenarios for the states along the

Mississippi-Missouri river corridor. Section 3 describes estimation issues and establishes the

econometric system. Section 4 presents the estimation results. Section 5 simulates crop

adaptation to climate change. Section 6 concludes.

2 Data

Geospatially explicit data on land cover, soil characteristics, weather, and climate change

scenarios are matched on a 4km by 4km grid to create the primary data set. The states

included in the analysis are those along the Mississippi-Missouri river corridor for which

there are at least 10 years of land cover data: Iowa, Illinois, Mississippi, and part of

Wisconsin, Missouri, and Arkansas. There is currently insufficient land cover data to extend

our analysis to other states. Summary statistics are provided in Table 1. Each variable in the

table is described in detail below.

Land use

Land cover data is derived from the Cropland Data Layer (CDL) available annually from

2000 to 2010 (USDA NASS) for the six states. The CDL is generated based on Resourcesat-

1 AWiFS, Landsat 5 TM, and Landsat 7 ETM+ satellites and has a ground resolution of 56 or

30 meters, depending on the year and sensors used (Mueller & Seffrin, 2006).

http://www.isro.org/

http://landsat.usgs.gov/

We divide land cover into major crops, other crops, non-crop and wild, urban, and

water bodies. The major crops include corn and soybean for Iowa, Wisconsin, and Illinois;

and corn, soybean, rice, and cotton for Missouri, Arkansas, and Mississippi. The category of

non-crop and wild land includes pasture, forest, improved pasture, etc. Conservation reserve

lands should fall within this category as they do not have crops. We define agricultural land

as the sum of major crops, other crops, and non-crop and wild land. Because urban and water

bodies are very difficult to convert into crop land, we do not include them in discussion in

this paper. Therefore, we define the share of major crops as the area of major crops divided

by area of agricultural land.

Figure 1 shows the shares of corn, soybean, rice, and cotton along the corridor. Corn

grows mainly in the colder north, while soy crops are more widely distributed. Rice and

cotton concentrate along the river in Missouri and Arkansas. For corn, the average percent

coverage from 2002 to 2010 is 34.3% in the north (the three northern states: Wisconsin, Iowa,

and Illinois), while it is only 2.9% in the south (the three southern states: Missouri, Arkansas,

and Mississippi). For soy, the coverage is 26.4% and 14.1% in the north and the south,

respectively. There is little cotton and rice in the north, while in the south, cotton takes 4.5%

of the agricultural land and rice takes 5%.

Soil Characteristics

For soil data we focus on two types of variables, both derived from the USDA’s U.S. General

Soil Map (STATSGO2). First, the underlying soil data include percent clay, sand, and silt,

water holding capacity, pH value, electrical conductivity, slope, frost-free days, depth to

water table, and depth to restrictive layer. Soil variable averages are spatially weighted from

irregular polygons for each grid cell.

Second, we use a classification system generated by the USDA – Land Capability

Class (LCC). A LCC value of one defines the best soil with the fewest limitations for

production, and progressively lower LCC classifications signify more limitations on the land

for agricultural production. The LCC integer scores decline incrementally to eight, where soil

conditions are such that agricultural planting is nearly impossible. The use of LCC codes add

explanatory power to the raw soil characteristics because these codes were assigned with

knowledge of past yields that depend on characteristics not present in our data set. The

distribution of LCC levels is shown in Figure 2. Together with Figure 1, we see that prime

agricultural soils are absent in southern Iowa and so largely is the corn-soy complex.

Similarly, more optimal soils hug the river in Missouri and Arkansas, and so do rice and

cotton.

Weather Variables

For weather data we use PRISM data processed by Schlenker and Roberts (2009) to a 4km by

4km spatial resolution, with a daily level of temporal resolution. The dataset includes both

temperature (highs and lows) and precipitation. Figure 3 shows the observed weather

condition in the planting season (from April to June)2 from 2002 to 2010 and the growing

seasons (from April to November) from 2002 to 2009. The observed temperatures are warmer

in the south and the precipitation levels are appreciably larger. Average temperature in the

growing season ranges from 12° to 25°c from the top of Iowa to the bottom of Mississippi, a

distance of 1600 km. Total rainfall in a growing season is also variable across this landscape

with a high of 130 cm and a low of 30 cm, highest in the southeast and lowest in the

northwest.

2 Planting season and growing season vary across crops. In the six states along the Mississippi-Missouri river corridor, the planting season is from April to May for corn, rice, and cotton, and from May to June for soybean. The harvest season is October for rice and corn, and November for cotton and soybean. Growing season is defined as the period between planting season and harvest season.

Because this study has so many cross-sectional data cells, we are able to use a great

deal of detail from the weather data. Two time periods of weather data are used for each crop

year. (1) The planting season data, which farmers know before they actually plant. A cold wet

spring, for instance, would delay planting and make a shorter season crop more desirable than

a longer season crop. Compared to corn, soy is more tolerant of being planted late and more

dependent on daylight hours, so it can make up time easily. When the planting season is late,

farmers are more inclined to plant soy. (2) Past weather is used as a proxy for expected

weather. We do not find much gain from including past weather beyond one season, though,

in terms of predicting current weather, quite a few lags of past weather are statistically

significant. For parsimony, we limit the lags of past weather to one.

Degree days are calculated from daily highs and lows using a fitted sine curve to

approximate the amount of hours the temperature is at or above a given threshold

(Baskerville & Emin, 1969). As in Schlenker and Roberts (2009), we bin the weather data

into degree days at a given temperature and above. We draw on their work and other

literature to reduce the number of bins to just those at critical thresholds. However, we

expand the number of classifications of temperature to account for the month in which it

occurs. We expect, for instance, that hot temperatures are not as harmful in autumn as they

are in the middle of the growing season.

Climate Change Scenarios

Climate change scenarios are taken from Climate Wizard. 3 Two models are considered: (1)

Ensemble average, SRES emission scenario: A1B; and (2) Ensemble average, SRES

emission scenario: A2. Both models predict temperature and precipitation in change and in

level for the end of the century (2080’s). The comparison baseline is the average temperature

and precipitation between 1961 and 1990. Future degree days are processed in two steps:

3 Source: http://www.climatewizard.org/

first, future temperature highs and lows are generated by adding changes to original highs and

lows; then the degree days are calculated based on the future highs and lows.

Figure 3 shows climate change scenarios, along with the observed weather condition

in the growing season in 2009 and the planting season in 2010. The observed temperatures

are warmer in the south and the precipitation levels are appreciably larger. Average

temperature in a growing season ranges from 12° to 25°c from the top of Iowa to the bottom

of Mississippi, a distance of 1600 km. Total rainfall in a growing season is also variable

across this landscape with a high of 130 cm and a low of 30 cm, highest in the southeast and

lowest in the northwest. The A1B model predicts a 4°c increase in temperature on average in

the north, and a 3.5°c increase in the south. The A2 model predicts a similar warming pattern,

but 0.5°c warmer than A1B’s prediction. The A1B model also predicts an 18 cm decrease in

total precipitation in a growing season in the north, and a 5 cm decrease in the south. The A2

model predicts a similar drying pattern with a very similar magnitude.

3 The Econometric System

Within each of our 4km grid cells, 𝑛, we observe the fraction of land in year 𝑡 that was

allocated to crop (or other use) 𝑖: 𝑆𝑖𝑛𝑡 . There are 𝑀 crops. If we imagine that each hectare of

our grid cells has a crop choice, then on that hectare the crop with the highest revenue will be

chosen. As a result, the fraction of the crop chosen in a grid cell will be a proportion type

model.

(1) 𝑆𝑖𝑛𝑡 = 𝜙�(𝜷𝟏′𝑿𝟏𝒏𝒕 + 𝑑1𝑛𝑡), … , (𝜷𝑴′𝑿𝑴𝒏𝒕 + 𝑑𝑀𝑛𝑡)�

where 𝑿𝒊𝒏𝒕 is a vector of determinate factors of revenue from planting crop 𝑖 on plot 𝑛 at year

𝑡, 𝜷𝒊 is a vector of coefficients and 𝑑𝑖𝑛𝑡 is an error term. 𝜙() is a suitable transformation with

its domain on the unit interval. When all of the shares are strictly within the unit interval,

using logit as the transformation and rearranging terms gives a linear estimation equation

(Berry 1994): 𝑙𝑜𝑔(𝑆𝑖𝑛𝑡) − 𝑙𝑜𝑔(𝑆0𝑛𝑡) = 𝜷𝒊′𝑿𝒊𝒏𝒕 + 𝑑𝑖𝑛𝑡. To deal with the fact that many plots

do not have a certain crop (i.e., many 𝑆𝑖𝑛𝑡 are zeros), we use a ratio transformation and we get

(2) 𝑆𝑖𝑛𝑡𝑆0𝑛𝑡

= 𝜷𝒊′𝑿𝒊𝒏𝒕 + 𝑑𝑖𝑛𝑡

In order to predict shares as a function of the independent variables, we sum the share

ratio over 𝑆𝑖 (recall that the shares sum to one) and solve for 𝑆0𝑛𝑡

(3) 𝑆0𝑛𝑡 = 11+∑ �𝜷𝒋′𝑿𝒋𝒏𝒕+𝑑𝑗𝑛𝑡�𝑀

𝑗=1

Substituting (3) into (2), we get

(4) 𝑆𝑖𝑛𝑡 = 𝜷𝒊′𝑿𝒊𝒏𝒕+𝑑𝑖𝑛𝑡1+∑ �𝜷𝒋′𝑿𝒋𝒏𝒕+𝑑𝑗𝑛𝑡�𝑀

𝑗=1

The estimation strategy is that first we estimate equation (2) by Tobit, accounting for

the zero shares. Then we simulate 𝑑𝑗𝑛𝑡 (𝑗 = 1, … ,𝑀) by taking draws from a left truncated

normal distribution with mean 0, standard deviation 𝜎𝑗𝑛𝑡 and truncation at −𝜷𝒊′𝑿𝒋𝒏𝒕. Finally,

we calculate 𝑆𝑖𝑛𝑡 for each draw and take the averages.

Because the scale of this study encompasses more than a thousand kilometers,

there are conditions that are unaccounted for in our variables that change across the

landscape. This spatial correlation can induce heteroscedasticity, which would make

straightforward tobit estimation inconsistent. We know of two feasible estimation

strategies. One strategy is to estimate a linear probability model with a Spatial Error Model

(SEM) correction for the errors. In the linear probability model, OLS would be consistent and

the SEM would serve to produce the correct standard errors and a more efficient estimate of

the coefficients. The limitation is that the prediction is not guaranteed to be between 0 and 1.

The other solution is to estimate local Tobit models, each for only one county and its

neighbors. The spatial correlation is taken care of because the coefficients and the variances

are free to vary across the landscape. Neighbors of county 𝑖 are defined to be counties whose

centroids are within 70 km distance of the centroid of county 𝑖. 70 km is chosen based on

Moran’s I tests. The tests show that the spatial correlation in error decrease exponentially and

beyond 70 km it is lower than 10−3. Within 70 km, a county has 8 neighbors on average and

each county has about 100 4km grid cells. Therefore, each regression has about 900

observations.

Next, we consider what explanatory variables should be included. The Nerlovian

adaptive price expectations model (Nerlove 1956) assumed that farmers have rational price

expectations based on their information set, and described it in three equations. Braulke

(1982) derived a reduced form from the three equations by removing the unobserved

variables. Choi and Helmberger (1993) combined this reduced form and farmer’s demand

functions, and based on their work, Huang and Khanna (2010) described the crop share as a

function of the lagged share, climate variables, economic variables, risk variables, population

density, and time trend. Hausman (2012) included most of these explanatory variables, and

also futures prices, substitute crop share and crop yield. To follow the literature,4 we include

lagged crop share, lagged substitute crop share, weather in the current planting season and the

last growing season, and soil conditions as explanatory variables. We include the interaction

term of heat and moisture to account for the possibility that dry warming is much more

harmful than warming with moisture (Lobell, et al. 2011) . We also include year fixed effects

to account for both output and input prices and government programs. This leads to the

following specification:

(5) 𝑆𝑖𝑛𝑡𝑆0𝑛𝑡

= 𝛼𝑖 + 𝛽𝑖𝑆𝑖𝑛𝑡−1 + 𝜸𝒊′𝑺𝑺𝑖𝑛𝑡−1 + 𝝋𝑖′𝑺𝒐𝒊𝒍𝑛 + 𝜽𝟏𝒊′ 𝑮𝑫𝑫𝑛𝑡−1 + 𝜽𝟐𝒊′ 𝑷𝑫𝑫𝑛𝑡 +

𝜽𝟑𝒊′𝑮𝑷𝑛𝑡−1 + 𝜽𝟒𝒊′ 𝑷𝑷𝑛𝑡 + 𝜽𝟓𝒊′ 𝑷𝒓𝒆𝑫𝑫𝑛𝑡+ 𝜽𝟔𝒊′ 𝑷𝒓𝒆𝑫𝑫𝑛𝑡−1 + 𝜇𝑡 + 𝜀𝑖𝑛𝑡

4 For reviews of share response literature, see Askari and Cummings (1977) and Nerlove and Bessler (2001) .

where 𝑆𝑖𝑛𝑡 is the share of crop 𝑖 planted at grid cell 𝑛 in year 𝑡 . 𝑺𝑺𝑖𝑛𝑡−1 is a vector of

substitute crop shares planted in year 𝑡 − 1. 𝑺𝒐𝒊𝒍𝑛 is a vector of soil conditions, including all

the soil characteristics described in the data section. 𝑮𝑫𝑫𝑛𝑡−1 is a vector of degree days by

month in the last growing season (April through November in year 𝑡 − 1). 𝑷𝑫𝑫𝑛𝑡 is a vector

of degree days by month in the current planting season (April through June in year 𝑡). The

critical temperatures in a planting season include 10 oc and 15 oc. 10 oc is the base temperature

limit of rice, corn, and soybean development, while 15 oc is the base temperature limit of

cotton development. The critical temperatures in a growing season include 10 oc, 15 oc, 20 oc,

25 oc, 29 oc, and 32 oc. Temperatures higher than 29 oc are harmful to corn, 30 oc to soybean,

and 32 oc to rice and cotton (Schlenker and Roberts 2009). 𝑮𝑷𝑛𝑡−1 is a vector of precipitation

by month in the last growing season. 𝑷𝑷𝑛𝑡 is a vector of precipitation by month in the current

planting season. 𝑷𝒓𝒆𝑫𝑫 are vectors of interactions of degree days above 30 oc and

precipitation levels in the same month. All months in the current planting season and the last

growing season are included.

4 Estimation Results

We run separate regressions for each crop and each county. In sum, we have 1022 sets of

estimates (368 counties; 2 main crops for the northern states and 4 main crops for the

southern states). We test the significance of soil, precipitation, and degree days. The F-test

results are shown in Table 2. Soil, precipitation and temperature are significant at the 1%

significance level in most of the regressions for corn, soy, and cotton, while they are

significant in half of the regressions for rice. Rice only covers about 4% of the land in the

southern states, while the land for other use covers about 80% of the land. It is not surprising

that the coefficients for rice are not statistically significant, given that the dependent variable

is the ratio of rice share and the share of other land use. In a linear probability model, using

just rice share, all coefficient groups are significant, so the lack of significance is likely

because of the inability to predict the “other” category. Cotton covers a small portion of land

as well, however cotton responds more strongly to weather than rice. Therefore, the

coefficients are significant in the regressions for cotton, while they are not in the regressions

for rice.

Based on the estimates, we predict crop share changes for two scenarios. In one

scenario, daily temperature increases by one degree for all months in 2009 and 2010. In the

other scenario, monthly precipitation decreases by one centimeter in all the months, and

temperature increases as above. We are interested in both short-run and long-run adaptation,

therefore we check the crop share changes in 2010, which is the year when the weather shock

happens, and in 2015, allowing the weather shock to take its full effect. The predicted crop

share changes are summarized in Table 3. In the short run, one-degree warming decreases

corn share by 0.007 in the north and increase corn share by 0.003 in the south, which means

0.7% less land (a 2% decrease) in the north and 0.3% more land (a 7% increase) in the south

is covered by corn. Although corn in the north and corn in the south are affected by warming

differently, corn in total is affected negatively, because it mainly grows in the north. One-

degree warming also decreases soy share in the north and increase soy share in the south. It

indicates that 2.1% less land (a 8% decrease) in the north and 2.8% more land (a 18.1

increase) in the south is covered by soy. One degree warming increases rice share by 0.031 (a

57.5% increase in the south) and cotton share by 0.023 (a 57.3% increase in the south). It

suggests that warming favors rice and cotton. By comparing the crop adaption in the north

and in the south, we find that both average shares of corn and soy decease in the north, while

all the main crop shares increase in the south. This finding contradicts the general hypothesis

that warming benefits the north agriculture. It suggests how warming affects crops depends

on more detailed weather and soil factors.

Compared to warming alone, dry warming increase other land use more in the north

and decrease other land use less in the south, as shown in Table 3. It shows that dry warming

does more damage to crop yields than warming with moisture. Although the averages are

different, the difference is small and the share change patterns are similar in the two

scenarios, as shown in Figure 4. This indicates that a one centimeter change in precipitation is

not large enough to have significant effects on crop adaptation.

Table 3 and Figure 4 also show the crop share changes in the long run. The crop share

changes in the long run are larger on average and the distributions have fatter tails. It suggests

that it takes time for farmers to fully adjust crop coverage to weather shocks. We also check

the crop share changes in 2020 and find that they are very similar to those in 2015. This

suggests that five years is long enough for the farmers to complete the adaptation.

To illustrate how crop adaptation varies across landscapes, we map out the long-run

share changes in Figure 5 and Figure 6 for the one-degree-warmer scenario and the one-

degree-warmer-and-one-centimeter-drier scenario, respectively. The findings are as follows.

First, the two scenarios have similar land cover shifting patterns, which confirms the findings

in Figure 4. Second, rice and cotton in the south spread toward the north, which is expected,

because the north becomes more suitable for rice and cotton. Third, the main crops take land

from minor crops and other uses in the south. This suggests that for south a one-degree

increase from current temperature is beneficial to the main crops. Finally, by comparing the

changing pattern of other land cover to the spatial distribution of LCC levels (Figure 2) and

precipitation (Figure 3 Panel B), we find that land with lower quality soils and more

precipitation are more likely to be converted into major crop land in face of climate warming.

To further investigate how soil affects crop adaptation, we construct a counterfactual

crop share change map for selected counties in Iowa. We choose one county in middle Iowa

and one in bottom Iowa according to their similarity in weather and their discrepancy in soil.

As shown in Figure 7, Panel A, in the growing season in 2009, the counties have similar

average temperatures which are around 14°c (14.0°c for middle Iowa and 14.7°c for bottom

Iowa) and similar precipitation levels, which are around 82 cm (82.7 cm and 82.1 cm,

respectively), while soils differ significantly (LCC level 2 vs. LCC level 3 and 6). Despite the

similar weather conditions and the same temperature increases, crop adaptations in the two

places are different. Changes in shares of corn, soybean, and other land use due to a one

degree increase in temperature are mapped out in the first and second row in Figure 7, Panel

B, for the middle Iowa county and the bottom Iowa county, respectively. The hypothesis is

that, if bottom Iowa had the same soil as middle Iowa, they would have similar crop

adaptation. To test this hypothesis, we predict the crop adaptation for the bottom Iowa

county assuming that they had the same soil as the middle Iowa county. First, we create the

counterfactual for the bottom Iowa county. We take the average soil properties (average LCC,

average percent of silt land, and averages of all other soil characteristics) of the middle Iowa

county, and the actual temperatures and precipitation levels of the bottom Iowa county.

Together they form the weather and soil conditions of the counterfactual land. Second, we

predict the crop shares for the counterfactual land. Two things are different from the

prediction for the actual bottom Iowa county – soils, and coefficients. Remember that

coefficients are changing across landscapes, because we run local regressions. The changing

coefficients reflect the fact that crops on landscapes with different soils are affected

differently by weather and soil. For example, precipitation on silt soil and sandy soil has

different effects on crop yields, because silt soil holds water more effectively than sandy soil.

The counterfactual has similar weather and soil to middle Iowa county, so we use the

coefficients estimated from the middle Iowa county to predict crop shares on the

counterfactual land. Next, we assume the temperature is one degree higher in all months in

the current planting season and the last growing season, and again predict the crop shares for

the counterfactual land. At last, we find the difference in the shares predicted from the last

two steps, and that is our predicted share change due to the one degree increase in

temperature. The results are shown in the last row of Figure 7, Panel B. Compared to the first

row of the same panel, it shows that the counterfactual land of the bottom Iowa county has

similar crop share change patterns as middle Iowa county. Figure 8 shows the distributions of

crop share changes for the middle Iowa county, the bottom Iowa county, and the

counterfactual land. This confirms the hypothesis above. The middle Iowa county and the

bottom Iowa county have different crop share change patterns. However, if the soils in the

bottom Iowa county were the same as those in the middle Iowa county, the crop share

changes would be similar to the changes in middle Iowa.

5 Climate Change Impacts

Given that farmers need about five years to fully adjust crop types to respond to a weather

shock, for the following discussion, we focus on crop share changes in the long run. Crop

adaptations under climate change are summarized in Table 4. Four climate change scenarios

are compared: (1) A warmer scenario predicted by the A1B model (only temperature changes

are considered), (2) A warmer-and-drier scenario predicted by the A1B model (both

temperature and precipitation changes are considered), (3) A warmer scenario predicted by

the A2 model, and (4) A warmer-and-drier scenario predicted by the A2 model. As shown in

Table 4, the four scenarios have similar effects on crop shares. In the north, the average

changes range from -0.0343 to -0.0514 for corn, from -0.0906 to -0.0986 for soy, and from

0.1249 to 0.15 for other land use. In the south, the average changes range from 0.0559 to

0.0702 for corn, from 0.1005 to 0.1118 for soy, from 0.0616 to 0.0714 for rice, from 0.0433

to 0.0572 for cotton, and from -0.2728 to -0.2956 for other land use.

The distributions of predicted crop share changes are depicted in Figure 9. Compared

to Figure 4, Figure 9 has wider distributions, which is expected because the A1B scenario has

larger increases in temperature than a one-unit increase. Spatial variations of crop adaptation

under the four scenarios are displayed out in Figures 10 through 13. Figure 10 considers

temperature changes only, predicted by the A1B model, while Figure 11 considers both

temperature and precipitation changes. The figures show similar land use shifting patterns,

which suggests that a drying climate within the predicted magnitude does not significantly

worsen the growth condition for crops. Therefore, we conclude that for the Mississippi-

Missouri river system, the major concern about climate change is warming, not drying.

Figures 12 and 13 consider the scenarios predicted by the A2 model. They are similar to

Figures 10 and 11, because the A2 model predicts the same patterns in temperature and

precipitation changes as the A1B model does, only with slightly larger magnitudes.

6 Conclusion

This paper examines crop adaptation to climate change in the context of the six states along

the Mississippi-Missouri river corridor. We consider the entire distribution of temperatures

within each day and each 4km grid cell. We also consider the soil conditions at the 4km grid

level. Based on the estimates of crop choices, we predict future crop share distribution under

several climate change scenarios. We find that rice and cotton spread north, while the average

shares of corn and soy decrease in the north and increase in the south. We also find that the

crop shifting pattern is not determined by temperature alone – soil plays an important role as

well, as there is less crop adaptation on prime soils than on lower quality soils. Therefore, due

to the variation in crop adaption on soils of varying quality, a significant makeover of major

crop distribution is not likely to happen.

Reference Askari, Hossein, and John T. Cummings. "Estimating Agricultural Supply Response with the Nerlove Model: a Survey." International Economic Review (257--292) 18, no. 2 (1977).

Baskerville, G.L., and P. Emin. "Rapid Estimation of Heat Accumulation from Maximum and Minimum Temperatures." Ecology, 1969: 514-517.

Berry, Steven T. "Estimating Discrete-choice Models of Product Differentiation." The RAND Journal of Economics, 1994: 242-262.

Braulke, Michael. "A Note on the Nerlove Model of Agricultural Supply Response." International Economic Review 23, no. 1 (1982): 241-244.

Chavas, Jean-Paul, and Matthew T. Holt. "Acreage Decisions Under Risk: the Case of Corn and Soybeans." American Journal of Agricultural Economics 72, no. 3 (1990): 529-538.

Choi, Jung-sup, and Peter G. Helmberger. "How Sensitive are Crop Yields to Price Changes and Farm Programs?" Journal of Agricultural and Applied Economics 25 (1993): 237-244.

Hausman, Catherine. "Biofuels and Land Use Change: Sugarcane and Soybean Acreage Response in Brazil." Environmental and Resource Economics 51, no. 2 (2012): 163-187.

Hausman, Catherine, Maximilian Auffhammer, and Peter Berck. "Farm Acreage Shocks and Food Prices: An SVAR Approach to Understanding the Impacts of Biofuels." Environmental and Resource Economics 53, no. 1 (2012): 117--136.

Huang, Haixiao, and Madhu Khanna. "An Econometric Analysis of US Crop Yield and Cropland Acreage: Implications for the Impact of Climate Change." AAEA annual meeting. Denver, Colorado, 2010. 25-27.

Just, Richard E. "An Investigation of the Importance of Risk in Farmers' Decisions." American Journal of Agricultural Economics 56, no. 1 (1974): 14-25.

Khanna, Madhu, Basanta Dhungana, and John Clifton-Brown. "Costs of Producing Miscanthus and Switchgrass for Bioenergy in Illinois." Biomass and Bioenergy 32, no. 6 (2008): 482--493.

Khanna, Madhu, Xiaoguang Chen, Haixiao Huang, and Hayri Onal. "Supply of Cellulosic Biofuel Feedstocks and Regional Production Pattern." American Journal of Agricultural Economics 93, no. 2 (2011): 473--480.

Lichtenberg, Erik. "Land Quality, Irrigation Development, and Cropping Patterns in the Northern High Plains." American Journal of Agricultural Economics 71, no. 1 (1989): 187-194.

Lin, William, and Robert Dismukes. "Supply Response Under Risk: Implications for Counter-cyclical Payments' Production Impact." Applied Economic Perspectives and Policy 29, no. 1 (2007): 64-86.

Lobell, David B., Marianne Banziger, Cosmos Magorokosho, and Bindiganavile Vivek. "Nonlinear Heat Effects on African Maize as Evidenced by Historical Yield Trials." Nature Climate Change 1, no. 1 (2011): 42-45.

McFadden, Daniel. "The Measurement of Urban Travel Demand." Journal of Public Economic 3, no. 4 (1974): 303−328.

McFadden, Daniel, et al. Demand Model Estimation and Validation. Vol. 4. Institute of Transportation Studies, 1977.

Miguez, Fernando E., Matthew Maughan, Germn A. Bollero, and Stephen P. Long. "Modeling Spatial and Dynamic Variation in Growth, Yield, and Yield Stability of the Bioenergy Crops Miscanthus x giganteus and Panicum Virgatum across the Conterminous United States." GCb Bioenergy, 2012: 509-520.

Miguez, Fernando E., Xinguang Zhu, Stephen Humphries, German A. Bollero, and Stephen P. Long. "A Semimechanistic Model Predicting the Growth and Production of the Bioenergy Crop Miscanthus x giganteus: Description, Parameterization and Validation." GCB Bioenergy 1, no. 4 (2009): 282-296.

Mueller, Rick, and Robert Seffrin. "New Methods and Satellites: A Program Update on the NASS Cropland Data Layer Acreage Program." Remote Sensing Support to Crop Yield Forecast and Area Estimates, ISPRS Archives 36, no. 8 (2006): W48.

Nerlove, Marc. "Estimates of the Elasticities of Supply of Selected Agricultural Commodities." Journal of Farm Economics 38, no. 2 (1956): 496-509.

Nerlove, Marc, and David A. Bessler. "Expectations, Information and Dynamics." Handbook of Agricultural Economics 1 (2001): 155-206.

Schlenker, Wolfram, and Michael J Roberts. "Nonlinear Temperature Effects Indicate Severe Damages to US Crop Yields under Climate Change." Proceedings of the National Academy of Sciences 106, no. 37 (2009): 15594-15598.

Schlenker, Wolfram, and Michael J. Roberts. "Nonlinear Temperature Effects Indicate Severe Damages to US Crop Yields under Climate Change." Proceedings of the National Academy of Sciences 106, no. 37 (2009): 15594-15598.

Schlenker, Wolfram, and Michael J. Roberts. "Nonlinear Temperature Effects Indicate Severe Damages to US Crop Yields under Climate Change." Proceedings of the National Academy of Sciences 106, no. 37 (2009): 15594-15598.

Scown, Corinne D., et al. "Corrigendum: Lifecycle Greenhouse Gas Implications of US National Scenarios for Cellulosic Ethanol Production." Environmental Research Letters 7, no. 1 (2012): 9502.

Taheripour, Farzad, Wallace E. Tyner, and MMichael Q. Wang. "Global Land Use Changes due to the US Cellulosic Biofuel Program Simulated with the GTAP Model." Argonne National Laboratory, http://greet. es. anl. gov/files/luc_ethanol, 2011.

Wu, JunJie, and Kathleen Segerson. "The Impact of Policies and Land Characteristics on Potential Groundwater Pollution in Wisconsin." American Journal of Agricultural Economics 77, no. 4 (1995): 1033-1047.

Figures 1: Observed Crop Coverage along the Mississippi-Missouri River System

Notes: Graphs display observed coverage shares for corn, soy, rice, cotton, and other land use, in the six states along the Mississipppi-Missouri river corridor. They are average shares over 2002-2010.

Figure 2: Distribution of Land Capabilty Classification (LCC) Levels

Notes: Land Capability Class (LCC) 1 is the best soil, which has the fewest limitations. Progressively lower classifications lead to more limited uses for the land. LCC 8 means soil conditions are such that agricultural planting is nearly impossible.

Figure 3: Observed Weather Conditions and Predicted Climate Change Scenarios Panel A. Temperature

Distribution is over 4km squares for temperature change to 2080.

0.5

11.

52

3 3.5 4 4.5 5 5.5Average Temperature Change

A1b A2

Panel B. Precipitation

Distribution is over 4km squares for precipitation change to 2080

0.0

2.0

4.0

6.0

8

-30 -20 -10 0 10Precipitation Change

A1b A2

Figure 4: Distribution of Crop Share Changes with Unit Change in Temperature and Precipitation

Notes: x-axes are crop share changes. For example, 0.2 in the first panel means corn share increases from 𝑎 to 𝑎+0.2. SR stands for Short Run, which is the year when the weather change happens. LR stands for Long Run, which is five years after the weather change happens. For corn and soy, all six states are included. For rice and corn, only the three south states are included, because there is no rice and cotton in the north.

02

46

810

-.6 -.4 -.2 0 .2 .4Corn

warmer SR warmer&drier SRwarmer LR warmer&drier LR

02

46

8

-.4 -.2 0 .2 .4 .6Soy


010

2030

40

-.5 0 .5 1Rice


05

1015

2025

-.5 0 .5 1Cotton


Figure 5: Crop Share Changes with Unit Increase in Temperature

Notes: a 20% change reported here means corn (for example) share increases from 𝑎 to 𝑎+0.2.

Figure 6: Crop Share Changes with Unit Increase in Temperature and Unit Decrease in

Precipitation


Figure 7: Counterfactual Analysis

Panel A. Similar in Weather and Different in Soil

Panel B. Crop Share Changes if Better Soil


Figure 8: Distributions of Crop Share Changes if Better Soil

Notes: x-axes are crop shares changes. For example, -0.1 in the first panel means corn share decreases from 𝑎 to 𝑎-0.1.

010

2030

4050

-.15 -.1 -.05 0 .05Corn

Middle Iowa Bottom IowaCounterfactual of Bottom Iowa

010

2030

-.1 -.05 0 .05 .1Soy


05

1015

20

-.1 0 .1 .2Other


Figure 9: Distribution of Predicted Crop Share Changes under Climate Change Scenarios

Notes: x-axes are crop share changes. For example, -0.5 in the first panel means corn share decreases from 𝑎 to 𝑎-0.5. For corn and soy, all six states are included. For rice and corn, only the changes in the three south states are included, because there is no rice and cotton in the north.

0.5

11.

52

-1 -.5 0 .5 1Corn

A1b Warmer A1b Warmer&DrierA2 Warmer A2 Warmer&Drier

0.5

11.

52

-.5 0 .5 1Soy


02

46

8

-.5 0 .5 1Rice


02

46

-.5 0 .5 1Cttn


Figure 10: Predicted Crop Share Changes under the A1B Scenario (Temperature Changes Only)


Figure 11: Predicted Crop Share Changes under the A1B Scenario (Temperature and

Precipitation Changes)


Figure 12: Predicted Crop Share Changes under the A2 Scenario (Temperature Changes Only)

Notes: a 20% change reported here means corn (for example) share increases from 𝑎 to 𝑎+0.2. .

Figure 13: Predicted Crop Share Changes under the A2 Scenario (Temperature and Precipitation

Changes)


Table 1: Summary Statistics North South Variable Mean St.dev. Min Max

Mean St.dev. Min Max

Dependent Variable: Percent Acreage (%) (Obs = 174825)

(Obs = 91584) Corn 0.343 0.179 0 0.983

0.029 0.067 0 0.774

Soy 0.264 0.142 0 1

0.141 0.189 0 0.955 Rice -- -- -- --

0.045 0.100 0 0.841

Cotton -- -- -- -- 0.050 0.122 0 0.982 Soil Condition (Obs = 19425)

(Obs = 10176)

Percent clay (%) 26.447 5.295 1.051 48.600

29.832 11.512 0 62.900 Percent sand (%) 21.560 12.642 0.793 95.300

26.942 14.235 0 84.600

Percent silt (%) 51.181 10.257 1.155 73.600

42.863 13.739 0 73.598 Water holding capacity 0.178 0.022 0.006 0.330

0.157 0.027 0 0.220

pH 6.441 0.511 0.198 7.700

5.591 0.686 0 7.511 Slope 3.435 4.536 0.037 48

7.203 6.394 0 25.400

Electrical conductivity 0.018 0.087 0 1

0.004 0.116 0 5.100 Frost free days 160.041 26.634 0 213.750

207.860 64.116 0 302.704

Depth to water table 79.960 41.173 0 201

62.777 22.750 13.250 201 Depth to restrictive layer 182.668 41.990 3.300 201

175.962 48.510 18 201

Percent of Land in Class 1 (%) 0.006 0.061 0 1

0.008 0.065 0 0.998 Percent of Land in Class 2 0.700 0.386 0 1

0.307 0.377 0 1

Percent of Land in Class 3 0.234 0.351 0 1

0.281 0.385 0 1 Percent of Land in Class 4 0.016 0.106 0 1

0.078 0.231 0 1

Percent of Land in Class 5 0.001 0.018 0 0.707

0.066 0.191 0 1 Percent of Land in Class 6 0.027 0.132 0 1

0.051 0.198 0 1

Percent of Land in Class 7 0.006 0.056 0 1

0.200 0.341 0 1 Percent of Land in Class 8 0 0 0 0 0 0.013 0 0.770

Table 1: Summary Statistics (continued) Weather Variables Planting Season (April through June from 2002 to 2010) (Obs = 174825)

(Obs = 91584)

Temperature (Daily Average, Celsius) 15.848 1.540 11.975 20.336

21.044 1.134 16.867 23.960 Precipitation (Total, CM) 32.678 9.969 5.640 74.760

36.311 5.018 16.410 51.130

Growing Season (April through November from 2002 to 2009 ) (Obs = 155400)

(Obs = 81408)


21.270 1.279 16.867 24.868 Precipitation (Total , CM) 74.022 14.814 32.540 127.290 86.839 9.109 56.730 120.550 Climate Change Scenarios (Obs = 19425)

(Obs = 10175)

A1B Planting Season (April through June) Temperature (Daily Average, Celsius) 4.381 0.340 3.868 5.544

3.609 0.261 3.016 4.095

Precipitation (Total, CM) -3.987 1.060 -7.518 -1.288

1.640 1.925 -3.346 5.533 Growing Season (April through November)


3.523 0.204 3.016 4.270 Precipitation (Total , CM) -17.592 4.334 -31.222 -6.461 -4.752 4.811 -23.613 6.639 A2

Planting Season (April through June) Temperature (Daily Average, Celsius) 5.018 0.385 4.393 6.281

4.096 0.263 3.533 4.602

Precipitation (Total, CM) -3.563 1.674 -8.547 -0.426

1.925 1.696 -2.058 5.969 Growing Season (April through November)


4.010 0.202 3.533 4.741 Precipitation (Total , CM) -17.794 4.106 -30.350 -6.087 -4.313 4.305 -21.719 8.605

Table 2: F-tests for Soil, Precipitation and Temperature

Regressions with 1% Significance Level Corn Soy Rice Cotton

Soil 93% 89% 54% 82%

Precipitation 76% 72% 45% 72%

Temperature 93% 90% 66% 92% Regressions with 5% Significance Level Corn Soy Rice Cotton

Soil 97% 92% 57% 85%


Temperature 95% 92% 67% 93% Regressions with 10% Significance Level Corn Soy Rice Cotton

Soil 97% 93% 60% 89%


Temperature 95% 93% 67% 93% Number of Regressions in Total 368 368 143 143

Table 3: Crop Share Changes with Unit Changes in Temperature and Precipitation

Corn Soy Rice Cotton Other

North Unit Changes Short-Run

Temperature Increase Only -0.0074 -0.0209 -- -- 0.0283

Temperature Increase and Precipitation Decrease -0.0067 -0.0257 -- -- 0.0324

Long-Run



Average Shares 0.3780 0.2623 0 0 0.3597

No. of Obs. 19425 19425 19425 19425 19425

South Unit Changes Short-Run

Temperature Increase Only 0.0026 0.0279 0.0307 0.0179 -0.0790

Temperature Increase and Precipitation Decrease 0.0044 0.0306 0.0225 0.0199 -0.0774

Long-Run



Average Shares 0.0349 0.1537 0.0534 0.0312 0.7268

No. of Obs. 10176 10176 10176 10176 10176

Notes: the numbers reported are share changes. For example, -0.0074 means corn share increases from 0.0378 (3.78% of land is covered by corn) to 0.3706.

Table 4: Crop Acreage Changes under Climate Change Scenarios

Corn Soy Rice Cotton Other North A1B Scenarios Long-Run



A2 Scenarios Long-Run



Average Shares 0.3780 0.2623 0 0 0.3597

No. of Obs. 19425 19425 19425 19425 19425 South A1B Scenarios Long-Run



A2 Scenarios Long-Run



Average Shares 0.0349 0.1537 0.0534 0.0312 0.7268

No. of Obs. 10176 10176 10176 10176 10176 Notes: the numbers reported are share changes.

Documents

Crop Adaptation to Climate Change - CERE- Centre for ... · Crop Adaptation to Climate Change 1 Introduction Crop yields are forecasted to decrease by 30-46% before the end of the